Cross-agent runtime compatibility test matrix

[tph-kds]

Context: manifests are generated for Claude (.claude-plugin/plugin.json), Codex (.codex-plugin/plugin.json), and Agents (.agents/plugins/marketplace.json) from config/skill-registry.json.

Gap: manifest validity is checked structurally, but no test verifies a DeckForge-produced application actually behaves correctly inside each agent runtime (tool discovery, skill loading, model routing).

Acceptance:

• Test matrix covering supported agent runtimes and their skill-discovery mechanics.
• At least one smoke test per runtime: load skill, generate a deck, run validators.
• Results recorded in the skill registry or a compatibility report; failures tracked per runtime.

[tph-kds]

Resolved. Delivered and merged via PR #23 (squash commit 22ca8b3) on 2026-08-09.

• Evidence: AGENT_RUNTIME_MATRIX.md + validate_runtime_matrix.py (commit 43fec73)
• Plan: docs/superpowers/plans/2026-08-07-backlog-resolution.md